Early Alzheimer’s Disease Detection Through Targeting the Feature Extraction Using CNNs

Alzheimer's Disease (AD) is a persistent, irreversible, and degenerative neurological disorder of the brain that currently has no effective therapy. This condition is identified by pathological abnormalities in the hippocampal area, which may develop up to 10 years prior to the onset of clinical symptoms. Timely detection of pathogenic abnormalities is essential to impede the worsening of AD. Recent studies on neuroimaging have shown that the use of Deep Learning techniques to analyze multimodal brain scans may effectively and correctly detect AD. The main goal of this work is to design and develop an Artificial Intelligence (AI) based diagnostic framework that can accurately and promptly detect AD by analyzing Structural Magnetic Resonance Imaging (SMRI) data. This study presents a novel approach that combines a Directed Acyclic Graph 3D-CNN with an SVM classifier for timely detection and identification of AD by analyzing the Regions of Interest (RoI) like cerebral spinal fluid, white and gray matter, and the hippocampus in SMRI images. The proposed hybrid model combines Deep Learning for feature extraction and Machine Learning techniques for classification. The obtained results demonstrate its superior performance compared to earlier methods in accurately identifying individuals with early mild cognitive impairment (EMCI) from those with normal cognition (NC) using the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset. The model attains a classification accuracy of 97.67%, with precision at 94.12%, and sensitivity at 98.60%.

@article{Prasad2025,
  title     = {Early Alzheimer’s Disease Detection Through Targeting the Feature Extraction Using CNNs},
  journal   = {International Journal of Advanced Computer Science and Applications},
  volume    = {16},
  number    = {1},
  year      = {2025},
  publisher = {The Science and Information Organization},
  author    = {D Prasad and K Jayanthi and Pradeep Tilakan},
  doi       = {10.14569/IJACSA.2025.0160157},
  url       = {https://doi.org/10.14569/IJACSA.2025.0160157}
}